Music and Data David Boyle interview with Alistair Croll

Hello.

Hey David, how are you?

Good. Good to speak to you. It's been a while.

Good to speak to you too. So thanks for talking to me. It's good to hear from you again and congratulations on landing somewhere interesting after emi.

Yeah, it's really good from that point of view. And I did. In between the two, I did HarperCollins, a book publishing. In my mind I'm exploring the similarities and differences between three industries where you've got creative people creating content and businesses putting out into the world. So I did four years as a, I guess as an executive in the fourth of the four record labels, but one that was under a huge amount of kind of commercial pressure thanks to being bought out by a venture capital company and then pressure to transform. So that was a. That was good because that really helped us sort lots of stuff out and then one that was ultimately sold. So it's packaged up and described and thought about a lot and then, and then streamlined and then sold. So there's actually quite a lot of helpful commercial pressures there which meant that we did some things other record labels didn't do. But there aren't that many people I think who I would recommend in terms of kind of data and music coming together, which I think is a kind of remarkable statement as well, given that I tried to try to seek them all out. There was. It was from a record label's point of view. So that's only one of many now I've left. I realize it's one of many kind of players in the industry. It seemed like it was the heart of the industry at the time, time. And I think in many ways it is. But if from a record label's point of view, it was really hard though, especially in. So it's 2009 to 2009. 10, 11, 12, 13. Yeah, 9, 10, 11, 12 and 13. In that time period it was really hard to get data. It was out there, it was definitely being generated, had been for years. But getting data as a record label in the value chain was really tough.

So why was that? It would seem to me like that would be the best place to get it.

Why is that? Yeah, so let me tell you what we had and then we can kind of work out. We didn't have as a result of that. So I mean we did have itunes. Transaction level records they give to everybody on every industry, which is helpful a little bit but you know, to understand. Well, we'll come on to that in a minute. So we did have that. Spotify was got up and running and Got got to some meaningful volume in that time period and they were very, very helpful at giving us more detailed transaction level records than itunes did and with more data about the individual. And of course there's more transactions in a streaming service than in the download service, so it's bigger volumes. That was pretty good, but outside of.

But presumably that's also less strong of a signal because someone isn't willing to put money behind it.

Yeah, exactly, exactly. But there were billions of records, so that was fun. But outside of that, I actually don't remember us having consumer research, which is the thing I spent most of my time on because it was the most valuable. We'll definitely talk about that. Outside of those three, I actually don't remember any other data sources playing a very meaningful role for us. A couple of caveats, I think. So one is that, you know, we were, we were interested in global stuff, so we're interested in data sets that worked across multiple countries. Lots of services, you know, Deezer at that time was big, but only in a small number, didn't touch the US and therefore wasn't a global supplier in that sense.

So what was the name of their organization?

Deezer.

What's Deezer?

Another streaming service like Spotify. So based out of France, they were getting to volume but not in, they were not in the UK for a lot of that time and they WEREN'T in the U.S. i think they now are in both. So Spotify was the only global streaming service to talk of. ITunes was the only meaningful.

Pandora didn't count.

No. I'm not sure if they, I can't remember. They didn't give us useful data or whether it wasn't sufficiently global enough. But either way it definitely wasn't on our radar.

And there's the irony that after Sean Fanning and Napster was unable to give us all the music we wanted for free, it went and hid in YouTube where it was the most inefficient place to give you music for free. And so there's a ton of music information on YouTube. I mean that messes things up, right? No one expected that to be where we get all our legal music from. So how did that affect things?

Absolutely, and Mark will talk more about that than I can. But yeah, I mean that definitely we saw it as a promotional platform. We saw, we looked at kind of play counts was our data set and we didn't get any meaningful data back that we analyzed at any scale or able to do so from them, despite it being the biggest we knew at the time. The biggest digital music consumption service. But I think what gets measured gets managed. And so. So we had itunes data and we had Spotify data. We had research, they got managed really nicely. But we didn't have a nice data set of YouTube plays and we didn't have a nice data set of all the other odds and ends. And so they really did get neglected as a result of that.

So are you using this analysis? There's sort of the. Tell me the known unknowns. Tell me things I already know, like what did we sell last month? Or how is Katy Perry doing? And then there's the unknown unknowns of like, which of these five acts is going to be big and we should sign them to the label. Where does that come down? Because it seems like most of your stuff was after the horse had left the barn.

Research. Yeah, it was. And it was for a couple of reasons I think though. I mean, one is that we wanted to be really sensitive about the role of data in a creative business, something I feel really strongly about. And so it's really clear that once you've recorded an album that everybody wants data to be used to help that album to find its audience. But it's not clear that it wasn't clear at all that people wanted to pick artists to sign based on it. So as that was kind of about political sensitivity as much as anything else. But then as we became as a function as we got a lot of trust and respect from people for. For the kind of post after all the stable optimization, so we went deeper down the value chain and there were some of the leading A and R people who used us to. To research both artists they hadn't signed yet. By. By. This is research. Right. So playing those songs to people around the world before they signed them and then. Yeah, and then giving us demo versions of songs to play before they'd committed to the final kind of edit and before they picked it as a single, before they put it on the album and given it to radio stations. So we did creep down the value chain towards the sensitive part in quite an important way with some really key people. But yeah, so I think it does.

Seem that the sort of the Simon Cowell manufactured boy bands thing was a precursor to that anyway. Right. It was just that Simon was a proxy for using data. But creating a boy band sounds entirely feasible out of data now.

I think that probably is more. Is much more feasible now. But it wasn't what we wanted to do. We wanted to be. We didn't want to create the perfect band. We wanted to help great talents to find their audience. So whether that was like, okay, we've signed you, but your audience is quite niche, we'll help you to find it, or whether it's, we've signed you, we wanted to make you huge. So it wasn't about the biggest and the best or it wasn't about creating. Creating that. So it's about what it is you want to achieve in the world.

Computers are for optimization, humans are for inspiration. But it seems to me like there's a real tension with that in the creative process.

Yeah, the same with the HarperCollins in book publishing and the same at BBC and TV. Absolutely. And all of those organizations have hits, have like very, very mainstream, you know, we might not be proud of it hits, but all of them pride themselves on a kind of portfolio. So, yeah, mainly it was after the horse had left and it was about which countries, which audiences for a particular artist and therefore which kind of marketing activities. That was the really the core of what we did. And then the other big area, particularly towards the end of that period when Spotify was establishing itself as a kind of global force, was analysis around what is the role of Spotify in people's lives? And, and so that's about what types of consumers is it engaging, what behaviors are they using it for relative to other services, and is it cannibalizing or additional to kind of consumption and engagement?

Do you think the labels cut Spotify more slack because they need its analytics?

No, I don't think so. They were not that smart, but now.

It seems like everybody needs the analysis. I was talking to Kato Reardon at Shazam and she said they can predict within five hours whether Katy Perry's song is going to beat Lady Gaga's just based on it's out for five hours, how many people ask what it is, and they give a pretty good estimate of how that song is going to do five months from now when it crosses the pond or something. It's crazy.

Yeah. I still don't think I would bet even today organizations as a whole aren't smart enough to factor that into the deal they do with the partner.

Yeah, I mean, people think that it's about where the revenue comes from, but it's much more about the insight that lets you adapt quickly.

Well, it shouldn't.

And that's a really big problem.

And it's really not music. And the other two, I've obviously advocated for the balance view and it's been really hard to translate that into the commercials of the deal.

So tell me For a second, if you would. You've got transactions from itunes, you've got transactions from Spotify. You said you have more information on the Spotify user. And so the metadata, I mean, just knowing transactions, that's just the sales report. That's for. That's accounting data. It's not that insightful. What is the meta like, where do you get good or bad metadata? What's the difference between those data sources?

Yeah, so itunes, so you get the transaction records, but with itunes, they give you partial zip code so you have some sense of geography. And yeah, that was it. Plus the anonymized user ID, which you can follow across other transactions. So you can look at lifetime engagement and. And if they downloaded the free song of the week, did they then go on to, you know, buy something? So that's at least a bit more interesting than transaction record. But Spotify would give a partial zip code as well, anonymized ID as well. But they would also give, I guess, metadata about the individuals. They would give gender and I think a rough age band. They would tell you whether it was mobile or desktop, whether they were premium or. Or free, the date of they first joined the service, a couple of other things like that. So it's a little bit more. They are basically trying to be helpful by giving us data that would help us to make decisions, but without revealing personal information. So it's a really nice balance of the two.

Yeah, it's tough because you kind of want to create Personas at least, but you have to make it from their anonymized data.

Yeah, so we did that towards the end of my time there. And we can show, for example, in the U.S. you know, we had a new artist and we could show the point at which they. They went from being. They started off being listened to by just a very niche audience genre in genre terms. And we could show the point at which they crossed into a more mainstream audience and so we could adapt our marketing spend accordingly. So stuff like that was. Which is actually completely revolutionary. You know, say on this week, the audience crossed over or you reached an inflection point. You know, it was completely revolutionary insight to see that on paper.

So how do change what you did when that tipping point happened?

Well, it's types of marketing, but more importantly, it's different communities. So a lot of you, you, you don't spend a lot of money on marketing and music until you really, really have to. So it would be niche marketing to communities through blogs and, you know, promotional opportunities. But as soon as you cross into the mainstream. That's when you have to A start spending money and B starting to get the music played in more mainstream places. So pulling the, pitching it to different record labels. Let's say we had a couple of those examples in the uk, all that be it. The UK examples were research based. So it's playing the song to people, which we did at great volume in the uk. There were two I can think of. So one was an artist that we'd signed believing he was going to be absolutely huge. But the research said no, no, no, he's not huge. It's a very niche appeal. It's actually young 13, 15 year olds. And so we reigned back our marketing spend and really held off at a big promotion until it was six months later he'd picked. So he got a number one single out of that because it was a niche audience that bought mus. Bought singles. Six months later there were two other very mainstream but old, sorry, very musicy but older segments that were pretty engaged. So we, so we open the tap a little bit more on the marketing. But it was a full year before the real really mainstream people were excited about his sound and we really put the mainstream marketing on. So it saved a ton of money on marketing for a year. And we didn't waste that marketing on an audience that wasn't ready for it, which we would have done otherwise. That would have been a disaster. And the flip side, there's another artist exactly the same time and we thought she was very niche. You can tell actually by the artwork on the album, it's just, it's aimed at a young, young audience. And in fact, right from pre release her music was incredibly appealing to very mainstream people. So we scrabbled to try to throw money, to pull money together from other pots for the marketing budget for this project because there wasn't any allocated and tried to invest in kind of mainstream promotional opportunities. And sure enough we sold a ton of albums which you really weren't expecting to do for, for a new artist. And I think, you know, if you look at the decision makers like let's say the head of marketing or something, they have so many personal opinions coming at them on a song or an artist. The radio stations, the buyers at the major retailers, the buyers at itunes, the promotional team at Spotify, they're all just maybe individually data driven, but they're all just kind of anecdotes by the time they hit that person's desk. And how the hell do you wrestle with all of that? So we would try to give that person Structure.

That's an interesting point, though. I remember the old days, it was like, you know, everyone's calling in, requesting that track. Who calls to request a track today? Like, you don't request a track, you go buy it on it on itunes or something. So, I mean, obviously itunes was a big shift. Amazon and Google are trying to Change that. There's YouTube, there's streaming services. It seems like the industry has gained new channels, but it's completely blind or relatively blind in those channels. Whereas someone like a Spotify knows a hell of a lot about who you are and what kind of music you want.

Yeah, absolutely. That's a great way to think of it. Let's just talk about TV for a second. Netflix is the kind of hottest property in tv, I think maybe in terms of a global impact, and they give broadcasters absolutely nothing. It's ridiculous. So by quarter, we get a number of viewers for each show. That is it. It's horrible.

For BBC stuff. For BBC stuff versus what you get on iplayer, which is incredibly rich.

Right, Exactly. Yeah. Yeah.

That's crazy.

So there's a trend, I think, across multiple industries towards. Yeah. Power being in the hands of a lot, often a lot of retailers instead of the.

Well, it's the last mile has the visibility, so it's the power. But if you look at Netflix and Amazon now, they're like, okay, we're studios and they have to be right. Otherwise an hbo, okay, we're going direct. So at what point does Spotify become a label?

Yeah, but I think Spotify, you know, they were. They were good people and they got. They got that we wanted to help grow the industry and grow our artists, and that was why, you know, we negotiated, but in a. It wasn't a tough negotiation to get data that we just talked about, but they were. They were really happy that, oh, you want this? What do you want it for? Oh, okay, great. Yeah, we'd love to help with that. Like, great. Because you know what? We want to understand and grow and promote our artists. We want to give you better artists for you to work with, and we want to understand the role of your service so we can help it to establish itself in the world. It was a great and really symbiotic relationship. So we told them things about their consumers that they didn't know. For example, we've been researching the service all around the world for a few years, and we talked to them about our different consumer segments. One of them was the casual kind of mainstream consumer, the big middle ground and their kind of head of marketing said, oh, okay, so can you tell us, has Spotify reached this, your casual consumer? And in which countries have they done so? Because they're all that's different in each country. And we said, sure. Like, here's all our research data, here's where we think you're mainstream, where we think your niche, you know. And then some of the analytics we were doing on. On usage of the Spotify platform was, was more than they were doing at the time. So we. So when I started, there were lots of. There were lots of kind of data sources lying around. The itunes stuff was lying around. It was literally just lying around. I mean, nobody was looking at it. We were lucky to have a guy in the IT department who thought it was important, had been gathering it because by the way, if you don't download your itunes transaction records, they go away. Luckily he'd been doing that and databasing it and it was on a plate when I started ready to be looked at. No one had. So there were things like that. And then we'd already done quite a lot of consumer research, spent hundreds of thousands of pounds on consumer research. And I wasted though my first few weeks looking at data we already had. And I say wasted because in the end all of it was very. Was very kind of biased in sample. So even our research that we sat on was amongst music consumers, which we were interested in much bigger picture issues to start with. Like all people and how they engage with music, not just music consumers. And all the digital digital services, of course, only had people who use their service and, you know, people buy from multiple services and you just can't join individuals across multiple services. So every data source we had had these huge, huge, huge, huge caveats around it. But we realized that now you can pretty easily gather pretty representative research in large quantities pretty effectively because you're presumably.

There'S also, you know, in the old days you'd have to go out and test stuff for the test audience. Now you can be like, come to this website, press play, tell us what you think. Like the whole thing can be automated, including auditioning a track. Right. And you can do a B testing of those tracks. And there's some. So, so the technology for that instead of like having you come in and I'm going to put a record on or something has changed.

Yeah. There's an important caveat there though, I think because lots of how music was tested in the US was like that. The, the recruitment methods were things on Facebook saying hey or whatever it was at the time saying hey, Come here and rate the latest new music. Right. But that attracts a certain type of person. So we were careful to use research agencies to find people so that they were admittedly people who were expecting a survey that day, but they didn't know whether it was washing powder or cars or music. So it's as unbiased as possible. Yeah. And we did a million of those interviews over just a couple of years. It's about $5 an interview.

Okay.

So. Yeah, not that much. Yeah.

That's not terrible, I guess, because you can automate stuff. Sure.

Yeah. And it's got 25 countries or so. We did that. Spread across. Yeah. Lots and lots of waves of research. Yeah. All their analysis was automated. So that's one of the differences as well. You know, where you do research, most people who do research on data, people. And so they take a long time writing a questionnaire. They do lots of very creative PowerPoint. At the end, we just automated everything. Data gathering, exercise. So as nationally representative as possible. Yeah. And you can ask about the suite of different behaviors, like iTunes behaviors versus Spotify versus CDs, versus magazines and online. And it's a little less accurate on each data point, but it's much richer. So. So, yeah, we started that with just one survey in one country. And even before we'd finished analyzing the results of that, it was so exciting. We did three more countries quickly, and then pretty soon we're doing 25 countries and regularly. And we were making really big decisions. But I think more exciting than the really big decisions we were making. Just hundreds and hundreds and hundreds, hundreds of decisions. The most exciting thing for me was that the data from that was given to people all around the organization and that the people at the grassroots were making little decisions every single day all around the world. As a result, anyone can pick an artist and predict whether they're going to do well or not. You gather lots of data, you do a special project, you hire some brains. Right. But to do that scale across the organization all around the world, to feed into the hundreds of little decisions that make that artist's career, that excites me much more.

Yeah. The democratization of it, rather than the one turn the ship kind of thing.

It was amazing. So every six months.

And so the artists had access to this data as well.

No, because the relationship with the artist was still very, very, very precious. And so it had to be kind of filtered before it went to the art.

Was that because you didn't want the artist to think you were pressuring them to change based on the data or just that this was a card that gave you some value in the relationship.

I think it just. It was a card with value, but the sensitivity of that was really important. So at the right point in each relationship, artists absolutely engaged with it. So, for example, I don't. Don't want to name names, but there was a major French DJ at the time who later became a worldwide dj, but at the time, he was a big French dj. We sat with him for three hours to talk through the research and the data, and he was elbowing me out the way so he could get close to the computer and read the data points off the graphs, that kind of thing. He was so engaged by it. And Stowe took it to heart to help him think about what he was. But everything from the music he was working on. So some genre information I was showing about how. And I made a joke about how people in Britain like 90s music and showed how relative to the rest of the world, this is a really strange British phenomenon. Just to illustrate to me, that was illustration of how the data had stories to make a wider point I was going to make. But he loved that from an A and R point of view and wanted to then do. He was smiling, so I said, why are you smiling? And he said, I'm going to bring that back. So he saw it as a challenge to overcome. Sure.

Because as a dj, you're trying to look different as well. So. And this is one of the interesting things is that, like, when everybody knows every topic that's safe with everyone else, the interesting topic is the one that nobody talks about. And so there's this weird. Like a classic example is like Avicii saying, hey, I'm gonna take country and show up at the Miami Music Festival and start doing like a folk song to beat. And everyone went. And there's nobody. Like everybody else who does it after him is derivative. There's just nobody else can claim that.

Right.

You bring a folk band on stage at the Miami festivals and that you're that. That guy. Right. Just like no one can wear a big mouse hat. I mean, and they're all looking for that thing. So I think the data can be useful not just to show you what people want, but what they don't expect. And I think in a world that's so sort of suffering from attention, poverty, novelty is a very important thing for artists, too.

So there are Quite a few DJs we worked with and then like Pink Floyd, the Beatles, and lots of big artists. So we would sit down with the Artist or with their management team and talked through it. But we did that. We used it to make quite big differences to how we worked with the artists.

So did any artists get really angry and say, you know, I'm not a machine, I am not a number, I'm a free man?

Absolutely many of them. But we just. But at the end of the day, the data was about understanding their audiences and understanding the opportunities to better engage their audiences. And so, you know, with one. What you want me to not name names? So you'd have to edit the video.

You don't, you don't have to name names.

It's okay. So with one, like household name, like rock artist with a long career. So we had a big row about one round a table with our chief executive and several of our board and there are about five or six representatives from the artist in question. And there was one opportunity which, which was so clear. So there's a segment of consumers who are really not engaged in music at all. Don't really, don't buy, buy much music, but really like this artist. And, and there was a real opportunity to, to kind of monetize that enthusiasm. But the only way that they were going to, they were going to pay for music for this artist was if it was very low price and if it was in places where they go to shop, which is not like, not very attractive places. Right. So we're saying make your music cheap. Put it in places that it's not stock today. And we were saying like, you know, don't put really nice packaging. They don't want really nice packaging. They want quite simple packaging. Okay. So simple packaging, low price. And in places where it's not stock today, like deeply unattractive.

Yeah, you're buying your, you're buying your Floyd album at Tesco or something. It's weird.

Yeah, well, that's not a bad example. And they, and where we got to was to say, okay, okay, okay, we get it, we see the opportunity, we see that that's what we would want to do. We just don't know if we want to go after it. Which I think was absolutely perfect. Right?

Well, yeah, because there's. Then that's a really good example. If you take the. Do your data. There's this line about data informed, not data driven.

Yeah.

And there's a case where you're being data informed. You're saying, we have made a decision to keep the brand elevated at the expense of what might be a short term gain that would cheapen the brand. And that's. That's a very, very difficult line to walk. But interesting that you're arming them with that. That's good.

And that would have been argued about for years. And in fact, the rest of the products that we were talking about in that meeting, mainly the rest of them were the high end, the deluxe products for the super engaged consumers, where we showed they wanted a lot more richness from the product suite, features and functions and extra content.

I love that you're calling this product suite with features and functions when what you're really saying is people with an Amex black card get to sit in an amphitheater and smell the guy's sweat while he's playing. You know?

Exactly. But they genuinely didn't know that their fans were willing to spend that much and wanted that much. And they genuinely thought before this process that if they did that they would be seen as being kind of terribly money grabbing. Yeah.

And I think there is the rise of the sort of super premium, you know, Roger Waters intimate venue, Paul McCartney come and have dinner. That kind of thing is there's definitely the stratification of what an artist can offer. And so on the one end, you have like this thing recently where Paul McCartney was on a song with, I think, Kanye and everyone's like, who's this McCartney guy? I love how Kanye is helping them to find him, really. But that's a great example. That's one end is like, put the guy out on a Kanye album. At the other end is you too can drink Sir Paul's sweat if you're close enough. Right. And there's. There's a whole. I don't know if that stratification would have been possible without data because. And without direct marketing. Right.

In this case, it absolutely would not. And so we argued about it. We kind of won the point, basically. Then, interestingly, as a record label, we went through a change of management. Our management team kind of wobbled on the research for. For a period of maybe six months. We got it back again. So there was this really fun point in that we were still negotiating the contract, renewing the contract with this band. So the band actually were then the strongest advocates of doing this suite of things that was outlined in the research. So the band held the record label in the contract to deliver what the research had said that the fans wanted, not the record label. It was brilliant. Came full circle.

But that's an interesting example because this is a band that's venerated. They've been around. Most of their albums are going to be reunion tours and Stuff I'm triangulating here, but at the same time you mentioned like the artists label for indie stuff. So if you're making cover art, I mean, who looks at cover art anymore? But let's say you're making some cover art, the design and the packaging of the artist, you know. Yes. You don't interfere with the creative process of the music. And so if Lord shows up and she's got a bunch of good songs and she's talented and so on, how long should her hair be like? Does that. You know, because there's the. There's the creation of the brand. That can be very data informed as well.

Yeah, we never went there either. Album art was actually pretty sacred as well, so that was all off limits. Unlike books, by the way, where they're totally happy to change absolutely anything about the COVID Totally sacred. So that didn't happen either.

And so much of what a musician does these days is driven by the other sources of revenue, merchandising, tours and so on. How much of the data would inform, you know, tour dates and things like that?

Well, generally the labels were not part of that.

But you think that would be vital information for someone who's then planning, you know, what size T shirt to make or where should the artist be stopping to play their first few shows.

Yeah, but again, generally the labels didn't have a cut of that and therefore it was handled in a way that wasn't very coordinated with the labels. So we absolutely offered to help with all that stuff. And there were definitely, you know, there were definitely artists I can think of. There's a famous. There was a trio of Swedish DJs we worked with who.

Mafioso, I think.

Exactly. So that some might say. So their manager really leapt on the kind of research and I'm pretty sure it was used to help Nicotine negotiate live deals.

Yeah, that makes sense. I mean, it seems like perfect information for that stuff.

It was just horribly uncoordinated because the labels didn't see that. You didn't see the end of recorded music and didn't get rights. They were music labels, not artist labels. So, yeah, not so much. We kind of brought together lots of this data. We did it in a way that was very democratizing and we achieved kind of quick, quite large grassroots change in the way people made decisions as a result, which was lovely. But then the end of that story is unfortunately tragic, which is that the label was then sold to Universal Music and. And it all decayed and crumbled and. And what if you look at Warner Music and Sony Music as well. All three, all three of the labels that now exist, none of the three labels that now exist have a program that's anywhere near as wide reaching or as bold or as accepted by the organization as what we'd achieved. So I think that the music industry actually took us, the recorded music industry, the record labels took a step back after the, after the merger. And I think if I look back on what it takes to really drive data, meaningful data into the decision making and democratize that across the business, I don't think that will emerge. And if I look at the kind of what's driving those businesses, the people who are driving them and the kind of pressures that are driving them, I don't think that they'll do that. So I think that the music industry will carry on being gut instinct led, anecdote driven, albeit sometimes with numbers attached to the anecdotes and it will continue to kind of make quite poor decisions. So I think that's the future of the recorded music industry because it's tough to bring all that data together, to tease out the right conclusions from it and to scale it across an organization. It's really tough. And the fact is this is something I really believe that the real decisions are made in an organization are kind of hundreds that add up to any one decision. And they're made in the corridors and in the lunchrooms and on the phone calls. These are not things that they're mostly not things people propose, prepare for or that you could even pitch your data towards. You have to get the data into the person so that they carry it with them. I think that's what we did and I think that's really tough. So I think you'll get great case studies, you'll get wonderful PowerPoints and you'll have some fantastic stories. But I think if you're looking at the change in the way a business operates, really across the organization, across countries and across departments, I think it will be really tough. That's my prediction. But two more thoughts then. So one is this economist I mentioned. So. So yeah, basically an economist, right? So he would look at different data sources and tease out inference and causation and that's all you need to know. You should definitely talk to him. We had him look at price optimization for us on itunes.

What was his name again?

Professor Dana Brett Danaher. I'll send you an introduction.

Oh sure, that'd be great. Thank you.

So for us at emi, we, he persuaded us to randomly allocate singles and albums, a selection of singles and albums to different price bands and just told us what the right answer was for pricing, which was not something that Apple agreed with. In fact, it's something they disagreed with. And we did it anyway. We made lots of money. So he so that's one kind of boring but interesting way we used him and it changed our entire pricing structure on digital and physical products. But then more interestingly though, we then had him look at itunes purchases in France when they introduced the Hadopi anti piracy law and controlling for trends in other countries, showing quite clearly that the announcement of the law, the publicity around the law did actually prevent piracy and drive people to purchase, which the government there didn't know at the time was quite a controversial political finding. That's a really interesting use of data to drive kind of quite high level decision making and lobbying and stuff.

This gives me like 50 things to think about. So thank you as always, David. It's fascinating. Next time we're in the same city, I lots of beers.

Great.

Thank you so much.

Thank you. Good talking to you later. Bye bye.